Computer Vision News 50 "Datasets through the LookingGlass" is a webinar series focused on reflecting in the data-related facets of Machine Learning (ML) methods. Our goal is to build a community of enthusiastic researchers interested who care about understanding the impact that data and ML methods could have in our society. The webinar is part of “Making MetaDataCount” project and is organized by Veronika Cheplygina (left in the picture) and Amelia Jiménez-Sánchez (on the right) at IT University of Copenhagen. We had five successful editions so far between 2023 and 2024 with 19 speakers in total, the videos are available on our YouTube playlist. Datasets through the L king-Glass In our last webinar one week ago (March 25), we covered several topics on dataset design and ML evaluation, including the importance of choosing the right metric for your analysis, which factors affect the creation of medical imaging datasets, and strategies to evaluate bias in skin lesion models. In this webinar, we had 5 excellent researchers providing insights about their work. ① Hubert Dariusz Zając and Natalia-Rozalia Avlona were the speakers of the first talk. Their research is a critical inspection of “ground truth” labels when working with medical imaging datasets. Hubert is a PhD student at the University of Copenhagen and a member of the Confronting Data Co-Lab, working on a healthcare AI project in Denmark and Kenya. Natalia is a lawyer and a Marie Curie PhD fellow (DCODE) at the Computer Science Department of the University of Copenhagen. Their curiosity in this topic was sparked by works highlighting that algorithms with high reported performances have been shown to suffer from shortcuts, i.e. spurious correlations between artifacts in images and diagnostic labels. Their aim was to address questions regarding who decides
RkJQdWJsaXNoZXIy NTc3NzU=